Some exciting news
Guidelight, or what I'm up to next
Tech
Clear-Eyed AI.
Steven Adler (ex-OpenAI) on AI capabilities, safety research, and what's needed for an exciting AI future.
Recent essays
30 of 46AI safety warnings are not marketing hype
I beg of you, please take AI companies seriously when they warn about the risks ahead
AI, compared to what?
Seven underappreciated ways to think about AI’s costs and benefits
The magic phrase that kills AI regulation
A "federal framework" isn't real AI policy unless it answers two questions
The magic phrase that kills AI regulation
A "federal framework" isn't real AI policy unless it answers two questions
The dawning of authoritarian AI
How the Pentagon is bullying AI companies into supporting mass surveillance and autonomous killing
The dawning of authoritarian AI
How the Pentagon is bullying AI companies into supporting mass surveillance and autonomous killing
Don't let AI companies grade their own homework
OpenAI's compliance with California law is questionable. Someone else should be checking.
Don't let OpenAI grade its own homework
OpenAI's compliance with California law is questionable. Someone else should be checking.
Judgment isn't uniquely human
Neither is taste. Why do we keep making this mistake?
Judgment isn't uniquely human
Neither is taste. Why do we keep making this mistake?
The phases of an AI takeover
Superintelligence might kill everyone on Earth. At least, that’s what the three most-cited AI scientists of all time believe.
The phases of an AI takeover
Superintelligence might kill everyone on Earth. At least, that’s what the three most-cited AI scientists of all time believe.
AI isn’t “just predicting the next word” anymore
On new abilities and how AI has changed
AI isn’t “just predicting the next word” anymore
On new abilities and how AI has changed
The skill I'd teach students for the AI era
Resolved: You should learn how to debate.
The skill I'd teach students for the AI era
Resolved: You should learn how to debate.
On erotica, mental health, and OpenAI's burden of proof
People deserve more than just a company’s word that it has addressed safety issues.
On erotica, mental health, and OpenAI's burden of proof
People deserve more than just a company’s word that it has addressed safety issues.
Five ways AI can tell you're testing it
Evaluation awareness, explained
Five ways AI can tell you're testing it
Evaluation awareness, explained
What I learned from the NYT's reporting on OpenAI's sycophancy crisis
New reporting on OpenAI's unused safety tools and how early the risks were known
What I learned from the NYT's reporting on OpenAI's sycophancy crisis
New reporting on OpenAI's unused safety tools and how early the risks were known
Around-the-clock intelligence
Today's AI is always available. What happens once it is always acting and always learning?
Around-the-clock intelligence
Today's AI is always available. What happens once it is always acting and always learning?
The 4.5 trillion dollar elephant in the room
Fear of potential NVIDIA retaliation is shaping important AI policy debates, whether NVIDIA means to or not
The 4.5 trillion dollar elephant in the room
Fear of potential NVIDIA retaliation is shaping important AI policy debates, whether NVIDIA means to or not
Practical tips for reducing chatbot psychosis
What AI companies can learn from a million-word delusion spiral
Practical tips for reducing chatbot psychosis
What AI companies can learn from a million-word delusion spiral
Review: If Anyone Builds It, Everyone Dies
On space probes, superintelligence, and engineering problems you absolutely have to get right